by Carlos Fernandes, Juan Laredo, Juan Merelo, Carlos Cotta, Agostinho Rosa in EvoPAR
This paper investigates dynamic and partially connected ring topologies for cellular Evolutionary Algorithms (cEA). We hypothesize that these structures maintain population diversity at a higher level and reduce the risk of premature convergence to local optima on deceptive and NP-hard fitness landscapes. A general framework for modelling partially connected topologies is proposed and three different schemes are tested. The results show that the structures improve the rate of convergence to global optima when compared to cEAs with standard topologies (ring, rectangular and square) on quasi-deceptive, deceptive and NP-hard problems. Optimal population size tests demonstrate that the proposed topologies require smaller populations when compared to traditional cEAs.
by Mario Garcia-Valdez, Juan-J. Merelo, Francisco Fernández de Vega in EvoAPPS posters
In this paper the effect of node unavailability in algorithms using EvoSpace, a pool-based evolutionary algorithm, is assessed. EvoSpace is a framework for developing evolutionary algorithms (EAs) using heterogeneous and unreliable resources. It is based on Linda’s tuple space coordination model. The core elements of EvoSpace are a central repository for the evolving population and remote clients, here called EvoWorkers, which pull random samples of the population to perform on them the basic evolutionary processes (selection, variation and survival), once the work is done, the modified sample is pushed back to the central population. To address the problem of unreliable EvoWorkers, EvoSpace uses a simple re-insertion algorithm using copies of samples stored in a global queue which also prevents the starvation of the population pool. Using a benchmark problem from the P-Peaks problem generator we have compared two approaches: (i) the re-insertion of previous individuals at the cost of keeping copies of each sample, and a common approach of other pool based EAs, (ii) inserting randomly generated individuals. We found that EvoSpace is fault tolerant to highly unreliable resources and also that the re-insertion algorithm is only needed when the population is near the point of starvation.
by Pablo García-Sánchez, Antonio Fernández-Ares, Antonio Miguel Mora, Pedro Ángel Castillo, Juan Julián Merelo, Jesús González
in EvoAPPS posters
This work presents the results obtained from comparing different tree depths in a Genetic Programming Algorithm to create agents that play the Planet Wars game. Three different maximum levels of the tree have been used (3, 7 and Unlimited) and two bots available in the literature, based on human expertise, and optimized by a Genetic Algorithm have been used for training and comparison. Results show that in average, the bots obtained using our method equal or outperform the previous ones, being the maximum depth of the tree a relevant parameter for the algorithm.
by Antonio J. Fernández Ares, Antonio M. Mora, Maribel García Arenas, Pablo García Sánchez, Juan Julian Merelo Guervós, Pedro A. Castillo in EvoAPPS posters
This paper presents an approach based in an evolutionary algorithm, aimed to improve the behavioralparameters which guide the actions of an autonomous agent (bot) inside the real-time strategy game, Planet Wars. The work describes a co-evolutionary implementation of a previously presented method(ANONYMOUSBot), which yielded successful results, but focused in 4 vs 4 matches this time. Thus, there have been analyzed the effects of considering several individuals to be evolved (improved) at the same time in the algorithm, along with the use of three different fitness functions measuring the goodness of each bot in the evaluation. They are based in turns and position, and also inmathematical computations of linear regression and area regarding the number of ships belonging to the bot/individual to be evaluated. In addition, the variance of using an evolutionary algorithm with and without previous knowledge in the co-evolution phase is also studied, i.e., respectively using specific rivals to perform the evaluation, or just considering to this end individuals in the population being evolved. The aim of these co-evolutionary approaches are mainly two: first, reduce the computational time; and second find a robust fitness function to be used in the generation of evolutionary bots optimized for 4 vs 4 battles.
by Victor Manuel Rivas Santos, Maria Isabel Garcia Arenas, Juan Julian Merelo Guervos, Antonio Mora Garcia and Gustavo Romero Lopezin EvoAPPS posters
by Federico Liberatore, Antonio Mora, Pedro Castillo, Juan Julián Merelo in EvoGAMES
Flocking strategies are sets of behavior rules for the interaction of agents that allow to devise controllers with reduced complexity that generate emerging behavior. In this paper, we present an application of genetic algorithms and flocking strategies to control the Ghost Team in the game Ms. Pac-Man. In particular, we define flocking strategies for the Ghost Team and optimize them for robustness with respect to the stochastic elements of the game and effectivity against different possible opponents by means of genetic algorithm. The performance of the methodology proposed is tested and compared with that of other standard controllers. The results show that flocking strategies are capable of modeling complex behaviors and produce effective and challenging agents.